55Wrapper to handle importing MPAS files into xarray.
66
77 Module:
8- 1. converts MPAS "xtime" to xarray time. Time dimension is assigned via
9- `preprocess_mpas(..., timeSeriestats=False, ...)`.
10- 2. converts MPAS "timeSinceStartOfSim" to xarray time for MPAS fields coming
11- from the timeSeriesStatsAM. Time dimension is assigned via
12- `preprocess_mpas(..., timeSeriesStats=True, ...)`.
13- 3. provides capability to remove redundant time entries from reading of
8+ 1. converts MPAS time in various formats to xarray time. The MPAS time
9+ variable is provided via
10+ `preprocess_mpas(..., timestr='xtime', ...)`.
11+ `timestr` can either be a single variable name or a pair of variable
12+ names. In the latter case, each time variable is converted to an
13+ xarray time and the mean of the two times is used as the final xarray
14+ time. Each variable name in `timestr` can refer either to a float
15+ array containing the the number of days since the start of the
16+ simulation (e.g. `daysSinceStartOfSim`) or a string variable with the
17+ date and time (e.g. `xtime`) in the usual MPAS format:
18+ YYYY-MM-DD_hh:mm:ss
19+ 2. provides capability to remove redundant time entries from reading of
1420 multiple netCDF datasets via `remove_repeated_time_index`.
1521
1622 Example Usage:
2127>>> ds = remove_repeated_time_index(ds)
2228
2329Phillip J. Wolfram, Xylar Asay-Davis
24- Last modified: 11/25 /2016
30+ Last modified: 12/01 /2016
2531"""
2632
2733import datetime
@@ -110,26 +116,44 @@ def ensure_list(alist): # {{{
110116 return alist # }}}
111117
112118
113- def time_series_stat_time ( timestr , daysSinceStart ): # {{{
119+ def get_datetimes ( time_var , yearoffset ): # {{{
114120 """
115- Modifies daysSinceStart for uniformity based on between differences
116- between MPAS-O and MPAS-Seaice.
121+ time_var is an xarray DataArray that contains time information as a date
122+ string, a floating-point number of days or a number of days represented
123+ as a pandas timedelta (in ns). The result is a list of datetimes
124+ corresponding to the input dates offset as appropriate by the yearoffset.
117125
118- Phillip J. Wolfram
119- 09/09 /2016
126+ Xylar Asay-Davis
127+ Last modified: 11/25 /2016
120128 """
121129
122- if (timestr == 'timeSeriesStatsMonthly_avg_daysSinceStartOfSim_1' ):
123- return [datetime .timedelta (x ) for x in daysSinceStart .values ]
130+ if time_var .dtype == '|S64' :
131+ # this is a variable like date strings like 'xtime'
132+ time = ['' .join (atime ).strip () for atime in time_var .values ]
133+ datetimes = [datetime .datetime (yearoffset + int (x [:4 ]), int (x [5 :7 ]),
134+ int (x [8 :10 ]), int (x [11 :13 ]),
135+ int (x [14 :16 ]), int (x [17 :19 ]))
136+ for x in time ]
137+ elif time_var .dtype == 'float64' :
138+ # this array contains floating-point days like 'daysSinceStartOfSim'
139+ start = datetime .datetime (year = yearoffset + 1 , month = 1 , day = 1 )
140+ datetimes = [start + datetime .timedelta (x )
141+ for x in time_var .values ]
142+ elif time_var .dtype == 'timedelta64[ns]' :
143+ # this array contains a variable like 'daysSinceStartOfSim' as a
144+ # timedelta64
145+ start = datetime .datetime (year = yearoffset + 1 , month = 1 , day = 1 )
146+ datetimes = [start + x for x in
147+ pd .to_timedelta (time_var .values , unit = 'ns' )]
124148 else :
125- return pd .to_timedelta (daysSinceStart .values , unit = 'ns' )
149+ raise TypeError ("time_var of unsupported type {}" .format (
150+ time_var .dtype ))
126151
127- # }}}
152+ return datetimes # }}}
128153
129154
130155def preprocess_mpas (ds , onlyvars = None , selvals = None , iselvals = None ,
131- timeSeriesStats = False , timestr = None ,
132- yearoffset = 1849 ): # {{{
156+ timestr = 'xtime' , yearoffset = 1849 ): # {{{
133157 """
134158 Builds correct time specification for MPAS, allowing a date offset because
135159 the time must be between 1678 and 2262 based on the xarray library.
@@ -141,50 +165,36 @@ def preprocess_mpas(ds, onlyvars=None, selvals=None, iselvals=None,
141165 conditions. Hence, a default date offset is chosen to be yearoffset=1849,
142166 (year 0001 of an 1850 run will correspond with Jan 1st, 1850).
143167
144- Note, for use with the timeSeriesStats analysis member fields set
145- timeSeriesStats=True and assign timestr.
146-
147- The timestr variable designates the appropriate variable to be used as the
148- unlimited dimension for xarray concatenation. For MPAS-O
149- timestr='time_avg_daysSinceStartOfSim' and for MPAS-Seaice
150- timestr='timeSeriesStatsMonthly_avg_daysSinceStartOfSim_1'.
168+ The data set is assumed to have an array of date strings with variable
169+ name given by timestr, typically either 'xtime', or 'xtime_start'.
151170
152171 The onlyvars option reduces the dataset to only include variables in the
153172 onlyvars list. If onlyvars=None, include all dataset variables.
154173
155174 iselvals and selvals provide index and value-based slicing operations for
156175 individual datasets prior to their merge via xarray.
157- iselvals is a dictionary, e.g., iselvals = {'nVertLevels': slice(0,3),
158- 'nCells': cellIDs}
159- selvals is a dictionary, e.g., selvals = {'cellLon': 180.0}
176+ iselvals is a dictionary, e.g. iselvals = {'nVertLevels': slice(0, 3),
177+ 'nCells': cellIDs}
178+ selvals is a dictionary, e.g. selvals = {'cellLon': 180.0}
160179
161180 Phillip J. Wolfram, Milena Veneziani, Luke van Roekel and Xylar Asay-Davis
162- 11/25/2016
181+ Last modified: 11/25/2016
163182 """
164183
165- # ensure timestr is specified used when timeSeriesStats=True
166- if timeSeriesStats :
167- if timestr is None :
168- assert False , 'A value for timestr is required, e.g., ' + \
169- 'for MPAS-O: time_avg_daysSinceStartOfSim, and ' + \
170- 'for MPAS-Seaice: ' + \
171- 'timeSeriesStatsMonthly_avg_daysSinceStartOfSim_1'
172-
173- # compute shifted datetimes
174- daysSinceStart = ds [timestr ]
175- datetimes = [datetime .datetime (yearoffset + 1 , 1 , 1 ) + x
176- for x in time_series_stat_time (timestr , daysSinceStart )]
184+ if isinstance (timestr , (tuple , list )):
185+ # we want to average the two
186+ assert (len (timestr ) == 2 )
187+ starts = get_datetimes (ds [timestr [0 ]], yearoffset )
188+ ends = get_datetimes (ds [timestr [1 ]], yearoffset )
189+ datetimes = [starts [i ] + (ends [i ] - starts [i ])/ 2
190+ for i in range (len (starts ))]
177191 else :
178- time = np .array (['' .join (atime ).strip () for atime in ds .xtime .values ])
179- datetimes = [datetime .datetime (yearoffset + int (x [:4 ]), int (x [5 :7 ]),
180- int (x [8 :10 ]), int (x [11 :13 ]),
181- int (x [14 :16 ]), int (x [17 :19 ]))
182- for x in time ]
192+ datetimes = get_datetimes (ds [timestr ], yearoffset )
183193
184194 assert_valid_datetimes (datetimes , yearoffset )
185195
186196 # append the corret time information
187- ds .coords ['Time' ] = pd . to_datetime ( datetimes )
197+ ds .coords ['Time' ] = datetimes
188198 # record the yroffset
189199 ds .attrs .__setitem__ ('time_yearoffset' , str (yearoffset ))
190200
0 commit comments